Desarrollo de un sistema de detección de personas usando inteligencia artificial para controlar el distanciamiento social en la zona comercial de Pelileo

This project aims to design a low-cost system that allows the detection of people with the use of artificial intelligence algorithms and measure the distance between these objects of study. This will be deployed in Pelileo specifically in the commercial area of Tambo to provide a method of crowd prevention. The second chapter contains a detailed description of the components that make up a digitised image, such as pixel, RGB space, colour map, etc.; as well as the image processing methods and artificial intelligence algorithms that allow the detection of objects, in this case people. The third chapter presents the description of the different components involved in the design and implementation of the system. First, the minimum specifications for the selection of the camera that meets the requirements of the system are analysed. An analysis of the artificial intelligence algorithms provided by Matlab is carried out in order to select the most suitable algorithm for the system, being ACF CALTECH the one with the best results. The calculation of the estimation of the distance between objects was implemented using test patterns with fixed distances in order to obtain data and calculate the equation that describes this estimation. A graphical explanation of the programming logic implemented is given by means of flow diagrams and finally, the graphical interface designed with the operation and the different modes that can be used is presented. Finally, the results obtained in the evaluation of the fidelity of the system are presented, where different tests were carried out at fixed camera distances and the ANOVA calculation was performed to determine whether the estimated distance differs when there is a change in the position of the object with respect to the camera. Tests were carried out to measure the error of the system under different fixed distances, obtaining an error rate of less than 12 %. Finally, the system was deployed in the commercial area of Pelileo -El Tambo and the operation of the system was checked, giving very satisfactory results in almost all conditions, having a recognition of people close to 100 % of effectiveness. Keywords: COVID-19, deep learning, images, videos, artificial intelligence, distance, pixels.

• Hlavní autor: Silva Chipantiza, Christian Fabricio (author)
• Médium: bachelorThesis
• Jazyk: spa
• Vydáno: 2023
• Témata: COVID-19; Aprendizaje profundo
• On-line přístup: http://dspace.unach.edu.ec/handle/51000/10688